scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-06 02:48:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 103

refactor(caching): remove redundant comments, restore warning-set state in test

Browse source
This commit is contained in:
Amr Mohammed 2026-09-27 16:30:31 +03:00
parent 1652079e21
commit c1f79fdf86
2 changed files with 13 additions and 24 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
litellm/caching/caching.py
View file

@ -394,10 +394,6 @@ class Cache:
param_value = kwargs[param]
cache_key += f"{param}: {param_value}"
elif kwargs[param] is not None and param not in _warned_dropped_cache_params:
# Provider-specific param (e.g. ollama num_ctx) is being excluded
# from the cache key. Two requests differing only in this param
# will collide and the second gets the first's cached response.
# Warn once per param; set enable_caching_on_provider_specific_optional_params=True to include it.
_warned_dropped_cache_params.add(param)
verbose_logger.warning(
"litellm.cache: provider-specific param '%s' is excluded from the cache key by default, "

33
tests/local_testing/test_caching.py
View file

@ -78,32 +78,25 @@ async def test_dual_cache_async_batch_get_cache():
def test_cache_key_warns_on_dropped_provider_specific_param(caplog):
"""Provider-specific params dropped from the cache key (default flag) should
emit a one-time warning. Regression for #42400."""
import litellm
from litellm.caching.caching import Cache, _warned_dropped_cache_params
litellm.enable_caching_on_provider_specific_optional_params = False
_warned_dropped_cache_params.discard("num_ctx") # reset in case another test warned
_warned_dropped_cache_params.discard("num_ctx")
cache = Cache()
try:
with caplog.at_level(logging.WARNING):
cache.get_cache_key(model="ollama/llama3.2",
messages=[{"role": "user", "content": "hello"}], num_ctx=2048)
assert any("num_ctx" in r.message for r in caplog.records)
with caplog.at_level(logging.WARNING):
cache.get_cache_key(
model="ollama/llama3.2",
messages=[{"role": "user", "content": "hello"}],
num_ctx=2048,
)
assert any("num_ctx" in r.message for r in caplog.records)
# second call must NOT warn again (one-time behavior)
caplog.clear()
with caplog.at_level(logging.WARNING):
cache.get_cache_key(
model="ollama/llama3.2",
messages=[{"role": "user", "content": "hello"}],
num_ctx=4096,
)
assert not any("num_ctx" in r.message for r in caplog.records)
caplog.clear()
with caplog.at_level(logging.WARNING):
cache.get_cache_key(model="ollama/llama3.2",
messages=[{"role": "user", "content": "hello"}], num_ctx=4096)
assert not any("num_ctx" in r.message for r in caplog.records)
finally:
_warned_dropped_cache_params.discard("num_ctx")
def test_dual_cache_batch_get_cache():